next up previous
Next: Online Annotation Up: Stroke Segmentation Previous: Combined Stroke Grouping and

The Approach Used by This System

Most of the methods considered for segmenting strokes into symbols did not appeal as they either relied on heuristics, or forced the user to have a certain behaviour.

If the character recogniser is able to return confidence information for interpretations of strokes, it is possible to automatically test different combinations of strokes and pick the best. This is a new approach, although is similar to that used by Yaeger, Webb and Lyon  which wasn't seen until afterwards.

The approach that the system uses is to combine overlapping strokes into indivisible units, then use the character recogniser to test groupings of strokes and these indivisible units. This solution works well, only limited by the strength of the underlying character recogniser. While it is less reliable than other stroke segmentation methods, such as the user pausing between characters, it does allow a much more natural and fluid entry of symbols into the system.

The process makes two assumptions: strokes that cross belong to the same character and all the strokes that belong to the one character will be drawn before the user moves onto the next. In other words, all i's must be dotted and all t's crossed before the next symbol is drawn. From casual observation of people writing with pen and paper, and from observation of people using this system, neither of these assumptions interfere with people's writing: most people tend to write like this anyway.

The process used to determine how to segment the strokes supplied by the user works as follows:

With the current character training data, k=4 so all groupings of 1, 2, 3, and 4 units are generated and tested. Due to the assumption that all strokes belonging to the same symbol are drawn in order, we can avoid a combinatorial explosion. The total number of combinations of up to the first k units is 2k - 1.

Figure 4.4: All groupings for four units.

Figure 4.4 illustrates the new method developed for generating all the groupings of up to four units. To generate all groupings of n units, the tree is built starting with a parent node of n 1's. The children of a node are then created by the addition of all adjacent pairs of boxes. Children that already exist elsewhere in the graph are not generated. The children that are not generated are represented by dotted boxes.

These groups indicate the groupings of strokes to try. For example, the group ``1 2 1'' means: ``Take the first unit on its own, then the next two units together, then the last unit on its own.''

next up previous
Next: Online Annotation Up: Stroke Segmentation Previous: Combined Stroke Grouping and
Steve Smithies